NASA Earth Exchange Global Daily Downscaled Projections (NEX- GDDP)

NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) 1. Intent of This Document and POC 1a) This document provides a brief overview of the NASA Earth Exchange (NEX) Global Daily Downscaled Projections (GDDP) dataset, and is intended for users who wish to apply the NEX- GDDP dataset in studies of climate change impacts. This document summarizes essential information needed for accessing and using information contained within the NEX-GDDP dataset. References and additional information are provided at the end of this document This NASA dataset is provided to assist the science community in conducting studies of climate change impacts at local to regional scales, and to enhance public understanding of possible future climate patterns at the spatial scale of individual towns, cities, and watersheds. This dataset is intended for use in scientific research only, and use of this dataset for other purposes, such as commercial applications, and engineering or design studies is not recommended without consultation with a qualified expert. Community feedback to improve and validate the dataset for modeling usage is appreciated. Email comments to bridget@climateanalyticsgroup.org and forrest.s.melton@nasa.gov. Dataset File Name: NASA Earth Exchange Global Daily Downscaled Projections (NEX- GDDP) Dataset URL: https://nex.nasa.gov/nex/projects/1356/ Data access URL: https://cds.nccs.nasa.gov/nex-gddp/ 1b) Technical points of contact for this dataset: Dr. Bridget Thrasher, bridget@climateanalyticsgroup.org Dr. Rama Nemani, rama.nemani@nasa.gov 2. Data Field Descriptions CF variable name, units: Spatial resolution: tasmin Daily Minimum Near-Surface Air Temperature Degrees Kelvin 0.25 degrees x 0.25 degrees 1

Temporal resolution and extent: Daily from 1950-01-01 00:00:00 to 2100-12-31 11:59:59 Units are in days since a reference date. The reference date varies by model, and is based on the reference date used in the corresponding CMIP5 GCM experiment. CF variable name, units: Spatial resolution: tasmax Daily Maximum Near-Surface Air Temperature Degrees Kelvin 0.25 degrees x 0.25 degrees Temporal resolution and extent: Daily from 1950-01-01 00:00:00 to 2100-12-31 11:59:59 Units are in days since a reference date. The reference date varies by model, and is based on the reference date used in the corresponding CMIP5 GCM experiment. CF variable name, units: Spatial resolution: pr Precipitation (mean of the daily precipitation rate) kg m-2 s-1 0.25 degrees x 0.25 degrees Temporal resolution and extent: Daily from 1950-01-01 00:00:00 to 2100-12-31 11:59:59 Units are in days since a reference date. The reference date varies by model, and is based on the reference date used in the corresponding CMIP5 GCM experiment. Dataset projection: Geographic Dataset datum: WGS-84 Location of pixel lat and lon Coverage: The pixel lat and lon fields in the metadata provide the location of the center of each pixel West Bounding Coordinate: 180 W East Bounding Coordinate: 180 E North Bounding Coordinate: 90 N South Bounding Coordinate: 90 S 2

3. Data Origin and Methods 3.1. Introduction The NEX-GDDP dataset is comprised of downscaled climate scenarios for the globe that are derived from the General Circulation Model (GCM) runs conducted under the Coupled Model Intercomparison Project Phase 5 (CMIP5) [Taylor et al. 2012] and across two of the four greenhouse gas emissions scenarios known as Representative Concentration Pathways (RCPs) [Meinshausen et al. 2011]. The CMIP5 GCM runs were developed in support of the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR5). This dataset includes downscaled projections from the 21 models and scenarios for which daily scenarios were produced and distributed under CMIP5. The purpose of these datasets is to provide a set of global, high resolution, bias-corrected climate change projections that can be used to evaluate climate change impacts on processes that are sensitive to finer-scale climate gradients and the effects of local topography on climate conditions. The demand for downscaling of GCM outputs arises from two primary limitations inherent with current global simulation results. First, most GCMs are run using relatively coarse resolution grids (e.g., a few degrees or 10 2 km), which limit their ability to capture the spatial details in climate patterns that are often required or desired in regional or local analyses. Second, even the most advanced GCMs may produce projections that are globally accurate but locally biased in their statistical characteristics (i.e., mean, variance, etc.) when compared with observations. The Bias-Correction Spatial Disaggregation (BCSD) method used in generating the NEX-GDDP dataset is a statistical downscaling algorithm specifically developed to address these current limitations of global GCM outputs [Wood et al. 2002; Wood et al. 2004; Maurer et al. 2008; Thrasher et al. 2012]. The algorithm compares the GCM outputs with corresponding climate observations over a common period and uses information derived from the comparison to adjust future climate projections so that they are (progressively) more consistent with the historical climate records and, presumably, more realistic for the spatial domain of interest. The algorithm also utilizes the spatial detail provided by observationally-derived datasets to interpolate the GCM outputs to higher-resolution grids. With the help of the computational resources provided by NEX and the NASA Advanced Supercomputing (NAS) facility, we have applied the BCSD method to produce a dataset of downscaled CMIP5 climate projections to facilitate the assessment of climate change impacts in the United States. The dataset compiles 42 climate projections from 21 CMIP5 GCMs (Table 1) and two RCP scenarios (RCP 4.5 and RCP 8.5) for the period from 2006 to 2100 (2 of the 21 models only provide data through 2099), as well as the historical experiment for each model for the period from 1950-2005. Each of these climate projections is downscaled at a spatial resolution of 0.25 degrees x 0.25 degrees (approximately 25 km x 25 km) resulting in a data archive size of more than 12 TB (1TB = 10 12 Bytes). This document provides a basic description of the implementation of the BCSD method as applied in the downscaling of the CMIP5 GCM data. Additional technical details for the algorithm may also be found in Wood et al. [2002, 2004], and Maurer et al. [2008]. The approach used to produce the NEX-GDDP dataset was previously applied to data from the CMIP3 archive, and the approach used in production of both datasets is described in detail in Thrasher et al. [2012]. 3

3.2 Methods 3.2.1 Datasets Climate Model Data: We compiled 42 climate projections from the 21 CMIP5 GCM simulations (Table 1) and two RCP scenarios (RCP 4.5 and RCP 8.5). Each of the climate projections includes daily maximum temperature, minimum temperature, and precipitation for the periods from 1950 through 2005 ( Retrospective Run ) and from 2006 to 2100 ( Prospective Run ). During the downscaling process, the retrospective simulations serve as the training data, and are compared against the observational climate records (see below). The relationships derived from the comparison are then applied to downscale the prospective climate projections. Because all 42 climate projections are downscaled through the same procedures, for simplicity we refer to them as GCM data without differentiating any individual models. Observational Climate Data: We use a climate dataset from the Global Meteorological Forcing Dataset (GMFD) for Land Surface Modeling, available from the Terrestrial Hydrology Research Group at Princeton University [Sheffield et al. 2006]. This dataset blends reanalysis data with observations and is currently available at spatial resolutions of 0.25 degrees, 0.5 degrees and 1.0 degree, and temporal resolutions of 3-hourly, daily, and monthly timesteps. For development of the NEX-GDDP dataset, we used the 0.25-degree, historical data for daily maximum temperature, daily minimum temperature, and daily precipitation from 1950 to 2005. 3.2.2 Data Pre-processing Since the BCSD method does not explicitly adjust the trends (the slopes, in particular) in climate variables produced by GCMs, we extract the monthly large-scale climate trends from the GCM temperature data. This is calculated as a 9-year running average for each individual month (e.g. the trend for all Januaries taken together). These trends are preserved and added back to the adjusted data after the bias-correction step. 3.2.3 Bias Correction (BC) The Bias-Correction step corrects the bias of the GCM data through comparisons performed against the GMFD historical data. For each climate variable in a given day, the algorithm generates the cumulative distribution function (CDF) for the GMFD data and for the retrospective GCM simulations, respectively, by pooling and sorting the corresponding source values (day of year +/- 15 days) over the period from 1950 through 2005. It then compares the two CDFs at various probability thresholds to establish a quantile map between the GCM data and the historical climate data. Based on this map, GCM values in any CDF quantile (e.g., p=90%) can be translated to corresponding GMFD values in the same CDF quantile. Assuming that the CDF of the GCM simulations is stable across the retrospective and the prospective periods, to correct the projected future climate variations the algorithm simply looks up the probability quantile associated with the predicted climate values from the estimated GCM CDF, identifies the corresponding observed climate values at the same probability quantile in the GMFD CDF, and then accepts the latter as the adjusted climate predictions. The climate projections adjusted in this way have the same CDF as the GMFD data; therefore, the possible biases in the statistical structure (the variance, in particular) of the original GCM outputs are removed by this procedure. 4

At the end of the Bias-Correction step, the previously extracted temperature climate trends are added back to the adjusted GCM climate fields 3.2.4 Spatial Disaggregation (SD) The Spatial-Disaggregation step spatially interpolates the Adjusted GCM data to the finer resolution grid of the 0.25-degree GMFD data. Other than simple linear spatial interpolation, multiple steps are adopted in the SD algorithm to preserve spatial details of the observational data. First, the multi-decade daily climatologies of the GMFD variables (temperature and precipitation) are generated at both native and GCM resolutions. The climatology for the SD step is the average for each day of the year calculated over the reference period, 1950-2005. Second, for each time step, the algorithm compares the Adjusted GCM variables with the corresponding GMFD climatology to calculate scaling factors. In particular, the scaling factors are calculated as the differences between the bias-corrected GCM and the GMFD data for temperature, but as the quotients (between the two datasets) for precipitation to avoid negative values for the latter. Third, the coarse-resolution scaling factors are bilinearly interpolated to the fine-resolution GMFD grid. Finally, the scaling factors are applied, by addition or shifting for temperatures and by multiplication for precipitation, on the fine-resolution GMFD climatologies to obtain the desired downscaled climate fields. As such, the algorithm essentially merges the observed historical spatial climatology with the relative changes at each time step simulated by the GCMs to produce the final results. 4. Considerations and Recommended Use 4.1 Recommended Use This dataset has been generated and is being distributed to assist the science community in conducting studies of climate change impacts at local to regional scales, and to enhance public understanding of possible future climate patterns and climate impacts at the scale of individual cities, communities, and watersheds. This dataset is intended for use in scientific research only, and use of this dataset for other purposes, such as commercial applications, and engineering or design studies is not recommended without consultation with a qualified expert. 4.2 Assumptions and Limitations The BCSD approach used in generating this downscaled dataset inherently assumes that the relative spatial patterns in temperature and precipitation observed from 1950 through 2005 will remain constant under future climate change. Other than the higher spatial resolution and bias correction, this dataset does not add information beyond what is contained in the original CMIP5 scenarios, and preserves the frequency of periods of anomalously high and low temperature or precipitation (i.e., extreme events) within each individual CMIP5 scenario. 4.3 Trend Adjustment to Individual Models As described in Section 2.1, the BCSD algorithm does not adjust the slope of the trends in the GCM projections. In the case of temperature, for instance, if the GCM predicts a mean temperature increase of 2 C between 2006 and 2100, the same temperature change (i.e., a trend 5

of 2 C over 95 years) will be observed in the downscaled temperature field. However, the BCSD algorithm does adjust the offset of the climate trends by shifting the retrospectively simulated climate variables (1950 through 2005) to match the GMFD data. In the previous example, if the simulated mean temperature from the GCM over the period 1996-2005 is 14 C, while the observed mean temperature is 15 C, the BCSD algorithm will correct the bias by shifting the GCM retrospective and prospective projections upward by 1 C. The adjusted mean temperature projected for the end of the 21 st century will then be raised from 16 C to 17 C, though its relative change over the period 2006-2100 is preserved as 2 C. Though such adjustments of future climate projections are qualitatively justifiable, quantitatively the linear shifting itself may not be realistic because the climate system is nonlinear in nature. Users of this dataset should be aware of this limitation of the downscaled data, particularly when using downscaled scenarios from individual GCMs. 5. Credits and Acknowledgements Please use the reference below as the primary citation for the methods used to produce this dataset: Thrasher, B., Maurer, E. P., McKellar, C., & Duffy, P. B., 2012: Technical Note: Bias correcting climate model simulated daily temperature extremes with quantile mapping. Hydrology and Earth System Sciences, 16(9), 3309-3314. Please add the following acknowledgement to any publications that result from use of this dataset: Climate scenarios used were from the NEX-GDDP dataset, prepared by the Climate Analytics Group and NASA Ames Research Center using the NASA Earth Exchange, and distributed by the NASA Center for Climate Simulation (NCCS). 6. References Daly, C., R.P. Neilson, and D.L. Phillips, 1994: A statistical-topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology, 33, 140-158. Maurer, E. P. and Hidalgo, H. G., 2008: Utility of daily vs. monthly large-scale climate data: an intercomparison of two statistical downscaling methods, Hydrology and Earth System Sciences, 12, 551-563. Meinshausen, M. S.J. Smith, K. Calvin, J.S. Daniel, M.L.T. Kainuma, and et al., 2011: The RCP greenhouse gas concentrations and their extensions from 1765 to 2300. Climatic Change, 109, 213-241. Sheffield, J., G. Goteti, and E. F. Wood, 2006: Development of a 50-yr high-resolution global dataset of meteorological forcings for land surface modeling, J. Climate, 19 (13), 3088-3111. 6

Shepard, D.S., 1984: Computer mapping: The SYMAP interpolation algorithm, in Spatial Statistics and Models, edited by G.L. Gaile and C.J. Willmott. D. Reidel Publishing Co., Norwell, Dordrecht, Holland, pp. 133-145. Taylor, Karl E., Ronald J. Stouffer, Gerald A. Meehl, 2012: An Overview of CMIP5 and the Experiment Design. Bull. Amer. Meteor. Soc., 93, 485 498. Thrasher, B., Xiong, J., Wang, W., Melton, F., Michaelis, A., & Nemani, R., 2013: Downscaled climate projections suitable for resource management. Eos, Transactions American Geophysical Union, 94(37), 321-323. Thrasher, B., Maurer, E. P., McKellar, C., & Duffy, P. B., 2012: Technical Note: Bias correcting climate model simulated daily temperature extremes with quantile mapping. Hydrology and Earth System Sciences, 16(9), 3309-3314. Wood, A.W., E.P. Maurer, A. Kumar, and D.P. Lettenmaier, 2002: Long-range experimental hydrologic forecasting for the eastern United States. J. Geophysical Research-Atmospheres, 107, 4429, doi:10.1029/2001jd000659. Wood, A.W., L.R. Leung, V. Sridhar, and D.P. Lettenmaier, 2004: Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs. Climatic Change, 15,189-216. 7. Dataset and Document Revision History Rev 0 08 June 2015 Document created. This is a new document/dataset. 7

Table 1. CMIP5 models included in GDDP dataset ACCESS1-0 CSIRO-MK3-6-0 MIROC-ESM BCC-CSM1-1 GFDL-CM3 MIROC-ESM-CHEM BNU-ESM GFDL-ESM2G MIROC5 CanESM2 GFDL-ESM2M MPI-ESM-LR CCSM4 INMCM4 MPI-ESM-MR CESM1-BGC IPSL-CM5A-LR MRI-CGCM3 CNRM-CM5 IPSL-CM5A-MR NorESM1-M 8